Transformer winding fault diagnosis method and device based on residual network

Through the residual network-based method, combined with current, temperature and vibration data, the modified residual network is used to diagnose transformer winding faults, which solves the problems of low efficiency, high cost and low accuracy of traditional detection methods, and achieves rapid Accurate fault identification and early warning ensure the stability of the power system.

CN118914927APending Publication Date: 2024-11-08STATE GRID XINJIANG ELECTRIC POWER CO LTD CHANGJI POWER SUPPLY CO
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411005597.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-25
Publication Date
2024-11-08

AI Technical Summary

Technical Problem

Traditional transformer winding fault detection methods are inefficient, costly and low accuracy, making it difficult to detect and diagnose faults in a timely and accurate manner, affecting the stability of the power system.

Method used

Using a residual network-based method, combining current, temperature and vibration data, through pre-processing, feature extraction, matrix construction and fault coefficient calculation, the pre-trained improved residual network is used for fault diagnosis, including introducing possible functions in the residual network. Deformed convolution and deformable spatial pyramid pooling layer.

Benefits of technology

It realizes rapid and accurate detection of transformer winding faults, improves detection accuracy and response speed, reduces costs, can identify faults in advance, reduce equipment losses, and ensures the reliability of power supply.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118914927A_ABST
    Figure CN118914927A_ABST
Patent Text Reader

Abstract

The invention provides a transformer winding fault diagnosis method and device based on a residual network, and belongs to the technical field of transformer fault classification detection, and the method comprises the steps: collecting transformer current data, transformer temperature data and transformer vibration data; the collected data are preprocessed; performing feature extraction on the preprocessed data; any two features are selected to construct a matrix; calculating a fault coefficient according to the norm of the matrix and a fault threshold; if the fault coefficient is not smaller than a preset value, the matrix is input into a pre-trained improved residual network, a transformer winding fault diagnosis result is obtained, and the pre-trained improved residual network comprises the steps that the residual network serves as a basic framework, deformable convolution is introduced into multiple layers or all layers of the residual network, and the fault diagnosis result of the transformer winding is obtained at the tail end of the residual network. And a deformable spatial pyramid pooling layer is adopted. The method is high in recognition accuracy, is small in error, and can find and recognize the fault of the transformer earlier.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of transformer fault classification detection, and in particular relates to a transformer winding fault diagnosis method and device based on a residual network. Background Art

[0002] As one of the indispensable and important equipment in the power system, the transformer plays a vital role in the power transmission and distribution process. The winding of the transformer is the core component of the transformer, and its operating status is directly related to the stability of the power system. However, since the winding may be affected by factors such as overload, short circuit, insulation aging, etc. during operation, the winding may fail. These failures will not only affect the normal operation of the transformer, but may also cause further damage to the equipment or even accidents. Therefore, it is very important to detect and diagnose the faults of the transformer winding in a timely and accurate manner. Traditional winding fault diagnosis methods mainly rely on manual inspection, offline detection and regular maintenance. This method is not only inefficient, but also difficult to detect and diagnose faults in a timely manner. Therefore, the development of an efficient and accurate transformer winding fault diagnosis method and device has become an urgent problem to be solved. With the development of the Internet of Things, big data and artificial intelligence technologies, various intelligent monitoring technologies have gradually received widespread attention. However, traditional intelligent transformer winding fault detection methods have problems such as low efficiency, high cost and low accuracy. Summary of the invention

[0003] In view of the shortcomings of the prior art, this application proposes a transformer winding fault diagnosis method and device based on a residual network. This application diagnoses transformer winding faults based on current, temperature, and vibration maps and a residual network, combines electrical signal analysis and deep learning technology, and has the advantages of high detection accuracy, fast response speed, and low cost, which is of great significance to improving the accuracy and reliability of transformer fault detection and diagnosis.

[0004] In a first aspect, the present application proposes a transformer winding fault diagnosis method based on a residual network, comprising:

[0005] Collect transformer current data, transformer temperature data and transformer vibration data;

[0006] Preprocessing the transformer current data, transformer temperature data and transformer vibration data;

[0007] Perform feature extraction on the preprocessed data;

[0008] Take any two features to construct a matrix;

[0009] Calculate the fault coefficient according to the norm of the matrix and the fault threshold;

[0010] If the fault coefficient is not less than a preset value, the data corresponding to the matrix is ​​abnormal. The matrix is ​​input into a pre-trained improved residual network to obtain a transformer winding fault diagnosis result. The pre-trained improved residual network includes: using a residual network as a basic architecture, introducing deformable convolutions in multiple layers or all layers of the residual network, and using a deformable spatial pyramid pooling layer at the end of the residual network.

[0011] The transformer current data acquisition process includes: for the high-voltage side winding of the transformer, using a current transformer to measure the transformer current data; for the low-voltage side winding of the transformer, using a Hall effect sensor to measure the transformer current data.

[0012] The transformer temperature data acquisition process includes: using a platinum resistance temperature sensor to measure the oil temperature of the transformer, and using an optical fiber temperature sensor to measure the winding temperature of the transformer.

[0013] The transformer vibration data acquisition process includes: measuring the transformer vibration data on the surface of the transformer winding using a piezoelectric acceleration sensor.

[0014] The matrix includes: a current and temperature characteristic matrix, a temperature and vibration characteristic matrix, and a vibration and current characteristic matrix.

[0015] The deformable convolution adjusts the position of the convolution kernel by learning the offset so that the convolution kernel is aligned to the key features in the matrix.

[0016] In a second aspect, the present application proposes a transformer winding fault diagnosis device based on a residual network, comprising: a sensor, a data acquisition card, a central processing module, and a fault analysis and diagnosis module; the sensor is connected to the data acquisition card, the data acquisition card is connected to the central processing module, and the central processing module is connected to the fault analysis and diagnosis module;

[0017] The sensor has a plurality of sensors, which are used to respectively collect transformer current data, transformer temperature data and transformer vibration data, and send the collected transformer current data, transformer temperature data and transformer vibration data to a data acquisition card;

[0018] A data acquisition card, used for converting the transformer current data, transformer temperature data and transformer vibration data into corresponding digital signals, and sending the digital signals to a central processing module;

[0019] A central processing module, used for storing historical digital signals and real-time digital signals, and sending the real-time digital signals to the fault analysis and diagnosis module;

[0020] The fault analysis and diagnosis module is used to preprocess the real-time digital signals corresponding to the transformer current data, the transformer temperature data and the transformer vibration data; perform feature extraction on the preprocessed data; select any two features to construct a matrix; calculate the fault coefficient according to the norm of the matrix and the fault threshold; if the fault coefficient is not less than a preset value, the data corresponding to the matrix is ​​abnormal, and the matrix is ​​input into a pre-trained improved residual network to obtain the transformer winding fault diagnosis result, wherein the pre-trained improved residual network includes: using the residual network as the basic architecture, introducing deformable convolution in multiple layers or all layers of the residual network, and using a deformable spatial pyramid pooling layer at the end of the residual network.

[0021] The sensors include: current transformer, Hall effect sensor, platinum resistance temperature sensor, optical fiber temperature sensor and piezoelectric acceleration sensor;

[0022] The current transformer is connected to the transformer to be detected and the data acquisition card through a shielded coaxial cable, and is used to detect the current of the high-voltage side winding of the transformer;

[0023] The Hall effect sensor is connected to the transformer and the data acquisition card respectively through a shielded coaxial cable, and is used to detect the current of the low-voltage side winding of the transformer;

[0024] The platinum resistance temperature sensor is connected to the transformer and the data acquisition card through insulated wires, and is used to measure the oil temperature of the transformer;

[0025] The optical fiber temperature sensor is connected to the transformer and the data acquisition card through insulated wires, and is used to detect the winding temperature of the transformer;

[0026] The piezoelectric acceleration sensor is connected to the transformer and the data acquisition card respectively through a plurality of pairs of shielded cables, and is used to measure the vibration data of the transformer.

[0027] In a third aspect, the present application proposes an electronic device comprising: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the transformer winding fault diagnosis method based on a residual network.

[0028] In a fourth aspect, the present application proposes a computer-readable storage medium storing executable instructions, which, when executed, enable a processor to execute the transformer winding fault diagnosis method based on a residual network.

[0029] Beneficial effects:

[0030] This application proposes a transformer winding fault diagnosis method and device based on a residual network. This application proposes a transformer winding fault diagnosis method based on current, temperature, vibration and residual network. The results show that this method has high recognition accuracy and small error. The test results show that this method can accurately detect and identify the type of transformer winding fault. It provides a new feasible solution for studying transformer winding fault identification. The fault diagnosis method of this application can detect and identify transformer faults earlier, provide early warning, minimize greater losses caused by major faults, and thus ensure the reliable supply of electricity. It has very important significance and practical value. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A flow chart of a transformer winding fault diagnosis method based on a residual network according to an embodiment of the present application;

[0032] Figure 2 Schematic diagram of feature extraction in an embodiment of the present application;

[0033] Figure 3 A schematic diagram of a fault determination process based on a fault coefficient according to an embodiment of the present application;

[0034] Figure 4 A schematic diagram of a transformer winding fault diagnosis device based on a residual network in an embodiment of the present application. DETAILED DESCRIPTION

[0035] The specific implementation of the present application is further described in detail below in conjunction with the drawings and examples.

[0036] This application proposes a transformer winding fault diagnosis method and device based on residual network to solve the problems of low efficiency, high cost and low accuracy in traditional transformer winding fault detection methods. Through signal acquisition, data processing, feature extraction, matrix construction, residual network training, fault diagnosis, result analysis and decision-making steps, a fast, reliable and accurate judgment of transformer winding faults can be achieved.

[0037] Embodiment 1:

[0038] This embodiment proposes a transformer winding fault diagnosis method based on a residual network, such as Figure 1 As shown, including:

[0039] Step S1: collecting transformer current data, transformer temperature data and transformer vibration data;

[0040] The transformer current data acquisition process includes: for the high-voltage side winding of the transformer, using a current transformer to measure the transformer current data; for the low-voltage side winding of the transformer, using a Hall effect sensor to measure the transformer current data.

[0041] The transformer temperature data acquisition process includes: using a platinum resistance temperature sensor to measure the oil temperature of the transformer, and using an optical fiber temperature sensor to measure the winding temperature of the transformer.

[0042] The transformer vibration data acquisition process includes: measuring the transformer vibration data on the surface of the transformer winding using a piezoelectric acceleration sensor.

[0043] In this embodiment, collecting transformer current data includes the following aspects:

[0044] (1) Choose a suitable current sensor:

[0045] For the high-voltage side winding, the LZZBJ9 current transformer is used to measure the current; for the low-voltage side winding, the A1120EUA Hall effect sensor can be used to measure the current. The Hall sensor can provide non-contact measurement and reduce the impact on the operation of the transformer.

[0046] (2) Installation location:

[0047] For the high voltage side, install the current transformer at the outlet of the high voltage side of the transformer; for the low voltage side, install the Hall effect sensor at the outlet of the low voltage side of the transformer.

[0048] (3) Connection:

[0049] Shielded coaxial cable with good electrical characteristics and shielding performance is used to ensure that the output signal line of the current sensor is connected to the data acquisition system to reduce electromagnetic interference.

[0050] (4) Data acquisition system:

[0051] A data acquisition card (DAQ) is used to collect the analog signals output by the sensor and convert them into digital signals.

[0052] (5) Data recording and storage:

[0053] Real-time monitoring: Real-time recording of current data to facilitate real-time monitoring of the transformer's operating status.

[0054] Historical data storage: The collected data is stored in the central processing module DSP28335 and used in the subsequent fault analysis and diagnosis module.

[0055] Through the above steps, the current data of the transformer winding can be effectively collected, and the current collected from the A phase of the transformer high-voltage side winding is IA = [IA1, IA2, IA3...IA 1000 ]; the current collected from the B phase of the transformer high voltage side winding is IB=[IB1,IB2,IB3......IB 1000]; the current collected from the C phase of the transformer high voltage side winding is IC=[IC1,IC2,IC3......IC 1000 ]; the current collected from the transformer low voltage side winding phase a is Ia=[Ia1,Ia2,Ia3......Ia 1000 ]; the current collected from the b phase of the transformer low voltage side winding is Ib=[Ib1,Ib2,Ib3......Ib 1000 ]; the current collected from the c phase of the transformer low voltage side winding is Ic=[Ic1,Ic2,Ic3......Ic 1000 ]. Provide important data support for transformer operation status monitoring and fault diagnosis.

[0056] In this embodiment, collecting transformer temperature data includes the following aspects:

[0057] The collection of transformer temperature is an important part of transformer status monitoring and fault prevention. Transformer temperature monitoring usually includes oil temperature monitoring and winding temperature monitoring.

[0058] (1) Selection of temperature sensor:

[0059] Oil temperature sensor: Use PT1000 platinum resistance temperature sensor to measure transformer oil temperature. This sensor has good stability and accuracy.

[0060] Winding temperature sensor: For monitoring the winding temperature, optical fiber temperature sensor can be used.

[0061] (2) Installation location:

[0062] Oil temperature sensor: Installed at the top and bottom center of the transformer tank to ensure that the oil temperature can be accurately reflected.

[0063] Winding temperature sensor: installed in the middle section of each phase winding of the transformer.

[0064] (3) Connection:

[0065] Use PTFE (polytetrafluoroethylene) insulated wire to ensure that the output signal line of the temperature sensor is connected to the data acquisition system. At the same time, pay attention to the stability of the connection to avoid poor contact due to vibration or thermal expansion and contraction.

[0066] (4) Data acquisition system:

[0067] Use a data acquisition card to collect the analog signal of the sensor, convert it into a digital signal and select a suitable sampling rate according to the speed of temperature change that needs to be monitored.

[0068] (5) Data recording and storage:

[0069] Real-time monitoring: Real-time recording of transformer temperature data to facilitate real-time monitoring of the transformer's operating status.

[0070] Historical data storage: The collected data is stored in the central processing module DSP28335, and these are used in the subsequent fault analysis and diagnosis module. Through the above steps, the temperature data of the transformer can be effectively collected. The temperature data of the top center of the transformer oil tank is TX = [TX1, TX2, TX3......TX 1000 ]; The temperature data of the center of the bottom of the transformer oil tank is TY=[TY1,TY2,TY3......TY 1000 ]; The temperature collected from the A phase of the transformer high voltage side winding is TA = [TA1, TA2, TA3......TA 1000 ]; the temperature collected from the B phase of the transformer high voltage side winding is TB = [TB1, TB2, TB3......TB 1000 ]; the temperature collected from the C phase of the transformer high voltage side winding is TC = [TC1, TC2, TC3......TC 1000 ]; the temperature collected from the transformer low voltage side winding phase a is Ta = [Ta1, Ta2, Ta3......Ta 1000 ]; the temperature collected from the transformer low voltage side winding phase b is Tb = [Tb1, Tb2, Tb3......Tb 1000 ]; the temperature collected from the c phase of the transformer low voltage side winding is Tc=[Tc1,Tc2,Tc3......Tc 1000 ]. Provide important data support for transformer operation status monitoring and fault prevention.

[0071] In this embodiment, collecting transformer vibration data includes the following aspects:

[0072] The collection of transformer winding vibration signals is an important part of transformer condition monitoring and fault diagnosis. The vibration signal can provide valuable information on the mechanical state and electrical performance of the winding.

[0073] (1) Selection of vibration sensor:

[0074] The piezoelectric PCB Piezotronics 352C23 accelerometer is selected because of its high sensitivity and can measure the vibration signal of the transformer more accurately.

[0075] (2) Installation location:

[0076] Install the vibration sensor on the surface of the transformer winding, at the top, middle and bottom of each phase winding of the transformer. The sensor should be firmly fixed on the winding to avoid loosening or falling off due to vibration.

[0077] (3) Connection and signal processing:

[0078] Use shielded multi-pair cables to connect the sensor and the data acquisition system to ensure that the output signal lines of the vibration sensor are connected to the data acquisition system to reduce electromagnetic interference.

[0079] (4) Data acquisition system:

[0080] Use a data acquisition card to collect the analog signal of the sensor and convert it into a digital signal. Select an appropriate sampling rate according to the frequency range of the vibration signal to ensure that the details of the vibration signal can be accurately captured.

[0081] (5) Data recording and storage:

[0082] Real-time monitoring: Vibration data is recorded in real time to facilitate real-time monitoring of the transformer's operating status.

[0083] Historical data storage: The collected data is stored in the central processing module DSP28335, and these are used in the subsequent fault analysis and diagnosis module. Through the above steps, the vibration signal of the transformer winding can be effectively collected. The vibration signal characteristics collected by the transformer are F = [F1, F2, F3......F 1000 ]; provide important data support for transformer operation status monitoring and fault diagnosis.

[0084] Step S2: preprocessing the transformer current data, transformer temperature data and transformer vibration data;

[0085] In this embodiment, the preprocessing includes the following contents:

[0086] (1) Data cleaning: Data cleaning is the first step in preprocessing, and its purpose is to remove irrelevant information, duplicate data, and erroneous data from the original data.

[0087] (2) Noise filtering: Due to various interference factors in the actual environment, vibration data often contains noise. The purpose of noise filtering is to remove these interferences and extract the real vibration signal.

[0088] (3) Data smoothing: Data smoothing is to eliminate short-term fluctuations and irregularities in the data and make the data smoother.

[0089] (4) Data synchronization and alignment: Ensure that data collected from different sensors are aligned in time for subsequent analysis.

[0090] Step S3: extracting features from the preprocessed data;

[0091] Feature extraction, such as Figure 2As shown in the figure, the following features are extracted: current waveform distortion, abnormal temperature rise, time domain features of vibration signals such as peak value, kurtosis, and waveform factor, and frequency domain features of vibration signals such as bandwidth, frequency component, and energy distribution. These features can reflect the health of the transformer.

[0092] Step S4: randomly select two characteristics to construct a matrix; the matrix includes: a current and temperature characteristic matrix, a temperature and vibration characteristic matrix, and a vibration and current characteristic matrix.

[0093] In this embodiment, the extracted feature data is constructed into a matrix form.

[0094] (1) The features of the current signal and the temperature signal are constructed into column vectors respectively, and then these column vectors are combined into a matrix. Each row of the matrix represents a feature vector at a time point.

[0095] R=[IA T , IB T ,IC T , Ia T , Ib T , Ic T , TX T , TY T , TA T , TB T , TC T ,Ta T , Tb T , Tc T ] (1)

[0096] The matrix is ​​composed of the characteristics of the current signal and the characteristics of the temperature signal. The current collected by the A phase of the transformer high-voltage side winding is IA = [IA1, IA2, IA3......IA 1000 ]; the current collected from the B phase of the transformer high voltage side winding is IB=[IB1,IB2,IB3......IB 1000 ]; the current collected from the C phase of the transformer high voltage side winding is IC=[IC1,IC2,IC3......IC 1000 ]; the current collected from the transformer low voltage side winding phase a is Ia=[Ia1,Ia2,Ia3......Ia 1000 ]; the current collected from the b phase of the low voltage side winding of the transformer is Ib=[Ib1,Ib2,Ib3......Ib 1000 ]; the current collected from the c phase of the transformer low voltage side winding is Ic=[Ic1,Ic2,Ic3......Ic 1000 ]; The temperature data of the top center of the transformer oil tank is TX=[TX1,TX2,TX3......TX1000 ]; the temperature data of the center of the bottom of the transformer oil tank is TY=[TY1,TY2,TY3......TY 1000 ]; The temperature collected from the A phase of the transformer high voltage side winding is TA = [TA1, TA2, TA3......TA 1000 ]; the temperature collected from the B phase of the transformer high voltage side winding is TB = [TB1, TB2, TB3......TB 1000 ]; the temperature collected from the C phase of the transformer high voltage side winding is TC = [TC1, TC2, TC3......TC 1000 ]; the temperature collected from the transformer low voltage side winding phase a is Ta = [Ta1, Ta2, Ta3......Ta 1000 ]; the temperature collected from the transformer low voltage side winding phase b is Tb = [Tb1, Tb2, Tb3......Tb 1000 ]; the temperature collected from the c phase of the transformer low voltage side winding is Tc=[Tc1,Tc2,Tc3......Tc 1000 ].

[0097] (2) The features of the current signal and the vibration signal are constructed into column vectors respectively, and then these column vectors are combined into a matrix. Each row of the matrix represents a feature vector at a time point.

[0098] V=[IA T , IB T ,IC T , Ia T , Ib T , Ic T , F T ] (2)

[0099] The matrix is ​​composed of the characteristics of the current signal and the characteristics of the vibration signal. The current collected by the A phase of the transformer high-voltage side winding is IA = [IA1, IA2, IA3......IA 1000 ]; the current collected from the B phase of the transformer high voltage side winding is IB=[IB1,IB2,IB3......IB 1000 ]; the current collected from the C phase of the transformer high voltage side winding is IC=[IC1,IC2,IC3......IC 1000 ]; the current collected from the transformer low voltage side winding phase a is Ia=[Ia1,Ia2,Ia3......Ia 1000 ]; the current collected from the b phase of the transformer low voltage side winding is Ib=[Ib1,Ib2,Ib3......Ib 1000]; the current collected from the c phase of the transformer low voltage side winding is Ic=[Ic1,Ic2,Ic3......Ic 1000 ]; the vibration signal characteristics collected by the transformer are F = [F1, F2, F3......F 1000 ].

[0100] (3) The features of the vibration signal and the temperature signal are constructed into column vectors respectively, and then these column vectors are combined into a matrix. Each row of the matrix represents a feature vector at a time point.

[0101] W=[ F T , TX T , TY T , TA T , TB T , TC T ,Ta T , Tb T , Tc T ] (3)

[0102] is a matrix composed of the characteristics of the vibration signal and the temperature signal. The vibration signal characteristics collected by the transformer are F = [F1, F2 F3......F 1000 ]; The temperature data of the top center of the transformer oil tank is TX=[TX1,TX2,TX3......TX 1000 ]; The temperature data of the center of the bottom of the transformer oil tank is TY=[TY1,TY2,TY3......TY 1000 ]; The temperature collected from the A phase of the transformer high voltage side winding is TA = [TA1, TA2, TA3......TA 1000 ]; the temperature collected from the B phase of the transformer high voltage side winding is TB = [TB1, TB2, TB3......TB 1000 ]; the temperature collected from the C phase of the transformer high voltage side winding is TC = [TC1, TC2, TC3......TC 1000 ]; the temperature collected from the transformer low voltage side winding phase a is Ta = [Ta1, Ta2, Ta3......Ta 1000 ]; the temperature collected from the transformer low voltage side winding phase b is Tb = [Tb1, Tb2, Tb3......Tb 1000 ]; the temperature collected from the c phase of the transformer low voltage side winding is Tc=[Tc1,Tc2,Tc3......Tc 1000 ].

[0103] Step S5: Calculate the fault coefficient according to the norm of the matrix and the fault threshold;

[0104] In this embodiment, different calculation formulas are used for different matrices, and the ratio between the norm of the matrix and the fault threshold is used to obtain the fault coefficient, which is described in detail as follows:

[0105] Construct the current signal and temperature signal matrix R, set a current and temperature fault coefficient A X , so that:

[0106]

[0107] Where x is the F-norm of the R matrix, x n The fault coefficient A of the current and temperature signal is X To preliminarily judge the transformer winding fault. When the current and temperature fault coefficient A X When it is not less than 1, it indicates that the current and temperature signals of the transformer winding are abnormal, and then the matrix is ​​input into the pre-trained improved residual network for further accurate judgment.

[0108] Construct the current signal and vibration signal matrix as V, set a current and vibration fault coefficient as B Y , so that:

[0109]

[0110] Among them, Y is the F-norm of the V matrix, Y n The fault coefficient B of the current and vibration signal is Y To preliminarily judge the fault of transformer winding. When the fault coefficient B of current and vibration Y When it is not less than 1, it indicates that the current and vibration signals of the transformer winding are abnormal, and then the matrix is ​​input into the pre-trained improved residual network for further accurate judgment.

[0111] Construct the vibration signal and temperature signal matrix as W, set a current and vibration fault coefficient as C Z , so that:

[0112]

[0113] Among them, z is the F-norm of the W matrix, z n The fault coefficient C of vibration and temperature is Z To preliminarily judge the transformer winding fault. When the fault coefficient C of vibration and temperature Z When it is not less than 1, it indicates that the vibration and temperature signals of the transformer winding are abnormal, and then the matrix is ​​input into the pre-trained improved residual network for further accurate judgment.

[0114] Step S6: If the fault coefficient is not less than the preset value, the data corresponding to the matrix is ​​abnormal, and the matrix is ​​input into the pre-trained improved residual network to obtain the transformer winding fault diagnosis result, which specifically includes: if the fault coefficient A of the current and temperature is X is not less than a preset value, then the data corresponding to the matrix is ​​abnormal, and the matrix is ​​input into the pre-trained improved residual network. At this time, the network will judge that the fault type of the transformer winding is the electrical contact fault and partial discharge fault of the winding; when the fault coefficient B of the current and vibration Y If the value is not less than the preset value, the data corresponding to the matrix is ​​abnormal. The matrix is ​​input into the pre-trained improved residual network. At this time, the network will judge that the fault type of the transformer winding is loose winding and inter-turn short circuit; the fault coefficient C of vibration and temperature Z If the value is not less than the preset value, the data corresponding to the matrix is ​​abnormal. The matrix is ​​input into the pre-trained improved residual network. At this time, the network will judge that the fault type of the transformer winding is the loose end clamp of the winding and the fatigue aging of the pad. The judgment flow chart is as follows: Figure 3 shown.

[0115] The pre-trained improved residual network includes: taking the residual network as a basic architecture, introducing deformable convolutions in multiple layers or all layers of the residual network, and adopting a deformable spatial pyramid pooling layer at the end of the residual network.

[0116] It is understandable that if the fault coefficient is less than the preset threshold, the data corresponding to the matrix does not have an abnormality, and the process ends directly without the need to perform transformer winding fault diagnosis based on the residual network. This embodiment preliminarily finds the corresponding abnormal data through the above judgment, which greatly reduces the amount of calculation and can meet the needs of real-time fault detection.

[0117] The deformable convolution adjusts the position of the convolution kernel by learning the offset so that the convolution kernel is aligned to the key features in the matrix.

[0118] In this embodiment, residual networks (Residual Networks, ResNets) are combined with deformable spatial pyramid pooling (Deformable Spatial Pyramid Pooling, DSPP) to construct a model that has both the feature extraction capability of a deep network and the ability to flexibly adapt to image feature representation capabilities of different scales and shapes, including the following:

[0119] (1) Build the residual network infrastructure:

[0120] First, build a residual network as the basic architecture. This network can be any standard residual network variant, such as ResNet-50, ResNet-101, etc. The residual network can learn rich feature representations through its residual blocks, and alleviate the gradient vanishing problem through skip connections, making the network deeper.

[0121] (2) Introducing deformable convolution:

[0122] Introduce deformable convolution in some or all layers of the residual network. Deformable convolution adjusts the position of the convolution kernel by learning the offset, so that the convolution kernel can be better aligned to the key features in the image, thereby improving the flexibility and accuracy of feature extraction.

[0123] (3) Apply deformable spatial pyramid pooling:

[0124] At the end of the residual network, Deformable Spatial Pyramid Pooling (DSPP) is applied. DSPP combines the concepts of deformable convolution and spatial pyramid pooling (SPP) to extract features of different scales by pooling the feature matrix at different scales. DSPP is able to generate a fixed-dimensional feature representation, which enables the model to be used for various image processing tasks such as classification, detection, etc.

[0125] During the training phase, standard back-propagation algorithms and gradient descent methods are used to optimize network parameters. Due to the combination of residual networks and DSPP, the network is able to learn richer and more flexible feature representations. During the optimization process, various techniques such as weight decay, learning rate scheduling, data enhancement, etc. can be used to improve the performance of the model.

[0126] Finally, the trained model is applied to the actual transformer winding fault diagnosis. Since the model combines the depth of the residual network and the flexibility of DSPP, it can achieve better performance.

[0127] By combining the residual network and DSPP, a model can be constructed that has both the feature extraction capability of a deep network and the ability to flexibly adapt to image feature representations of different scales and shapes. This combination has broad application prospects in various visual tasks.

[0128] Embodiment 2:

[0129] This embodiment proposes a transformer winding fault diagnosis device based on a residual network, such as Figure 4 As shown, it includes: a sensor, a data acquisition card, a central processing module, and a fault analysis and diagnosis module; the sensor is connected to the data acquisition card, the data acquisition card is connected to the central processing module, and the central processing module is connected to the fault analysis and diagnosis module;

[0130] The sensor has a plurality of sensors, which are used to respectively collect transformer current data, transformer temperature data and transformer vibration data, and send the collected transformer current data, transformer temperature data and transformer vibration data to a data acquisition card;

[0131] A data acquisition card, used for converting the transformer current data, transformer temperature data and transformer vibration data into corresponding digital signals, and sending the digital signals to a central processing module;

[0132] A central processing module, used for storing historical digital signals and real-time digital signals, and sending the real-time digital signals to the fault analysis and diagnosis module;

[0133] The fault analysis and diagnosis module is used to preprocess the real-time digital signals corresponding to the transformer current data, the transformer temperature data and the transformer vibration data; perform feature extraction on the preprocessed data; select any two features to construct a matrix; calculate the fault coefficient according to the norm of the matrix and the fault threshold; if the fault coefficient is not less than a preset value, the data corresponding to the matrix is ​​abnormal, and the matrix is ​​input into a pre-trained improved residual network to obtain the transformer winding fault diagnosis result, wherein the pre-trained improved residual network includes: using the residual network as the basic architecture, introducing deformable convolution in multiple layers or all layers of the residual network, and using a deformable spatial pyramid pooling layer at the end of the residual network.

[0134] The sensors include: current transformer, Hall effect sensor, platinum resistance temperature sensor, optical fiber temperature sensor and piezoelectric acceleration sensor;

[0135] The current transformer is connected to the transformer to be detected and the data acquisition card through a shielded coaxial cable, and is used to detect the current of the high-voltage side winding of the transformer;

[0136] The Hall effect sensor is connected to the transformer and the data acquisition card respectively through a shielded coaxial cable, and is used to detect the current of the low-voltage side winding of the transformer;

[0137] The platinum resistance temperature sensor is connected to the transformer and the data acquisition card through insulated wires, and is used to measure the oil temperature of the transformer;

[0138] The optical fiber temperature sensor is connected to the transformer and the data acquisition card through insulated wires, and is used to detect the winding temperature of the transformer;

[0139] The piezoelectric acceleration sensor is connected to the transformer and the data acquisition card respectively through a plurality of pairs of shielded cables, and is used to measure the vibration data of the transformer.

[0140] Embodiment 3:

[0141] This embodiment proposes an electronic device, including: one or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the transformer winding fault diagnosis method based on the residual network.

[0142] The electronic device may be a mobile phone, a computer or a tablet computer, etc., including a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, a transformer winding fault diagnosis method based on a residual network as described in the embodiment is implemented. It can be understood that the electronic device may also include an input / output (I / O) interface and a communication component.

[0143] The processor is used to execute all or part of the steps in the transformer winding fault diagnosis method based on the residual network as described in the above embodiment. The memory is used to store various types of data, which may include instructions of any application or method in the electronic device, as well as data related to the application.

[0144] The processor can be an application specific integrated circuit (ASIC), a digital signal processor (DSP), a programmable logic device (PLD), a field programmable gate array (FPGA), a controller, a microcontroller, a microprocessor or other electronic components, and is used to execute a transformer winding fault diagnosis method based on a residual network described in the above embodiment.

[0145] Embodiment 4:

[0146] This embodiment provides a computer-readable storage medium storing executable instructions. When the instructions are executed and implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0147] The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of a transformer winding fault diagnosis method based on a residual network described in each embodiment of the present application.

[0148] The aforementioned storage media include: flash memory, hard disk, multimedia card, card-type memory (for example, SD (Secure Digital Memory Card) or DX (Memory Data Register, MDR abbreviation, memory data register) memory, etc.), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, disk, CD, server, APP (Application, abbreviation of application software) application store and other media that can store program verification codes, on which computer programs are stored. When the computer program is executed by the processor, the various steps of the above-mentioned transformer winding fault diagnosis method based on the residual network can be implemented.

[0149] The various embodiments in the present application are described in a progressive manner, and the same or similar parts between the various embodiments can be referenced to each other, and each embodiment focuses on the differences from other embodiments.

[0150] The protection scope of the present application is not limited to the above-mentioned embodiments. Obviously, those skilled in the art can make various changes and modifications to the present disclosure without departing from the scope and spirit of the present disclosure. If these changes and modifications fall within the scope of the claims of the present disclosure and their equivalents, the intention of the present disclosure also includes these changes and modifications.

Claims

1. A transformer winding fault diagnosis method based on residual network, characterized in that: include: Collect transformer current data, transformer temperature data and transformer vibration data; Preprocessing the transformer current data, transformer temperature data and transformer vibration data; Perform feature extraction on the preprocessed data; Take any two features to construct a matrix; Calculate the fault coefficient according to the norm of the matrix and the fault threshold; If the fault coefficient is not less than a preset value, the data corresponding to the matrix is ​​abnormal. The matrix is ​​input into a pre-trained improved residual network to obtain a transformer winding fault diagnosis result. The pre-trained improved residual network includes: using a residual network as a basic architecture, introducing deformable convolutions in multiple layers or all layers of the residual network, and using a deformable spatial pyramid pooling layer at the end of the residual network.

2. The transformer winding fault diagnosis method based on residual network according to claim 1 is characterized in that: The transformer current data acquisition process includes: for the high-voltage side winding of the transformer, using a current transformer to measure the transformer current data; for the low-voltage side winding of the transformer, using a Hall effect sensor to measure the transformer current data.

3. The transformer winding fault diagnosis method based on residual network according to claim 1 is characterized in that: The transformer temperature data acquisition process includes: using a platinum resistance temperature sensor to measure the oil temperature of the transformer, and using an optical fiber temperature sensor to measure the winding temperature of the transformer.

4. The transformer winding fault diagnosis method based on residual network according to claim 1 is characterized in that: The transformer vibration data acquisition process includes: measuring the transformer vibration data on the surface of the transformer winding using a piezoelectric acceleration sensor.

5. The transformer winding fault diagnosis method based on residual network according to claim 1 is characterized in that: The matrix includes: a current and temperature characteristic matrix, a temperature and vibration characteristic matrix, and a vibration and current characteristic matrix.

6. The transformer winding fault diagnosis method based on residual network according to claim 1 is characterized in that: The deformable convolution adjusts the position of the convolution kernel by learning the offset so that the convolution kernel is aligned to the key features in the matrix.

7. A transformer winding fault diagnosis device based on residual network, characterized in that: include: Sensor, data acquisition card, central processing module, fault analysis and diagnosis module; the sensor is connected to the data acquisition card, the data acquisition card is connected to the central processing module, and the central processing module is connected to the fault analysis and diagnosis module; The sensor has a plurality of sensors, which are used to respectively collect transformer current data, transformer temperature data and transformer vibration data, and send the collected transformer current data, transformer temperature data and transformer vibration data to a data acquisition card; A data acquisition card, used for converting the transformer current data, transformer temperature data and transformer vibration data into corresponding digital signals, and sending the digital signals to a central processing module; A central processing module, used for storing historical digital signals and real-time digital signals, and sending the real-time digital signals to the fault analysis and diagnosis module; A fault analysis and diagnosis module is used to pre-process the real-time digital signals corresponding to the transformer current data, transformer temperature data and transformer vibration data; The preprocessed data is subjected to feature extraction; two features are randomly selected to construct a matrix, and the fault coefficient is calculated according to the norm of the matrix and the fault threshold; if the fault coefficient is not less than a preset value, the data corresponding to the matrix is ​​abnormal, and the matrix is ​​input into a pre-trained improved residual network to obtain a transformer winding fault diagnosis result, wherein the pre-trained improved residual network comprises: taking the residual network as the basic architecture, introducing deformable convolution in multiple layers or all layers of the residual network, and adopting a deformable spatial pyramid pooling layer at the end of the residual network.

8. The transformer winding fault diagnosis device based on residual network according to claim 7 is characterized in that: The sensors include: current transformer, Hall effect sensor, platinum resistance temperature sensor, optical fiber temperature sensor and piezoelectric acceleration sensor; The current transformer is connected to the transformer to be detected and the data acquisition card through a shielded coaxial cable, and is used to detect the current of the high-voltage side winding of the transformer; The Hall effect sensor is connected to the transformer and the data acquisition card respectively through a shielded coaxial cable, and is used to detect the current of the low-voltage side winding of the transformer; The platinum resistance temperature sensor is connected to the transformer and the data acquisition card through insulated wires, and is used to measure the oil temperature of the transformer; The optical fiber temperature sensor is connected to the transformer and the data acquisition card through insulated wires, and is used to detect the winding temperature of the transformer; The piezoelectric acceleration sensor is connected to the transformer and the data acquisition card respectively through a plurality of pairs of shielded cables, and is used to measure the vibration data of the transformer.

9. An electronic device, characterized in that: include: One or more processors, and a memory, wherein the memory is used to store instructions, and when the instructions are executed by the one or more processors, the one or more processors execute the transformer winding fault diagnosis method based on the residual network as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that: It stores executable instructions, which, when executed, enable a processor to execute the transformer winding fault diagnosis method based on a residual network as claimed in any one of claims 1 to 6.